aDepartamento de Genética, Instituto de Biociências, Universidade Federal do Rio Grande do Sul, Caixa Postal 15053, 91501-970, Porto Alegre, RS, Brazil
bLaboratório de Bioinformática, Modelagem e Simulação de Biossistemas, Faculdade de
Informática, Pontifícia Universidade Católica do Rio Grande do Sul, Av. Ipiranga 6681, 90610-001, Porto Alegre, RS, Brazil
Keywords: Glycolytic proteins; Molecular modeling, Protein structure, Alcohol
dehydrogenase, Functional Diversification, Divergence, Enzyme Function, ADH, Molecular evolution
*Corresponding author. Tel.: 55 51 3217-4182; Fax: 55 51 3308-9823; E-mail address: [email protected]
57 The plant alcohol dehydrogenases (ADHs) have been intensively studied in the last years in terms of phylogeny and they have been widely used as a molecular marker. However, almost no information about their three-dimensional structure is available. Several studies point to functional diversification of the ADH, with evidence of its importance, in different organisms, in the ethanol, norepinephrine, dopamine, serotonin, and bile acid metabolism. Computational results demonstrated that in plants these enzymes are submitted to a functional diversification process, which is reinforced by experimental studies indicating distinct enzymatic functions as well as recruitment of specific genes in different tissues. The main objective of this article is to establish a correlation between the functional diversification occurring in the plant alcohol dehydrogenase family and the three-dimensional structures predicted for 17 ADH belonging to Poaceae, Brassicaceae, Fabaceae, and Pinaceae botanical families. Volume, molecular weight and surface areas are not markedly different among them. Important electrostatic and pI differences were observed with the residues responsible for some of them identified, corroborating the function diversification hypothesis. These data furnish important background information for future specific structure-function and evolutionary investigations.
58
Introduction
The alcohol dehydrogenase (ADH) proteins belong to the medium-chain dehydrogenases/reductases (MDR) superfamily which has almost 1000 members spread in all types of organisms. MDR-ADH have been described in bacteria, archaea, yeast, plants and animals, and have been additionally implicated in ethanol oxidation, norepinephrine, dopamine, serotonin and bile acid metabolism (Höög et al. 2001), as well as in the in vitro and in vivo oxidation of retinol (Boleda et al. 1993; Martras et al. 2004).
Alcohol dehydrogenases (ADHs) are dimeric enzymes of the glycolytic pathway which encode two types of enzymes, one characterized by short protein chains (~250 residues), represented by Drosophila ADHs, which do not require zinc as a cofactor; and another characterized by long protein chains (~370 residues), represented by ADHs from organisms as diverse as mammals, plants and yeasts, which require zinc as a cofactor, and are called class P in dicot and monocot plants. The highest specificity of the ADHs among the latter is for ethanol, aldehyde, and acetaldehyde substrates, but they can also utilize other primary alcohols as well (Garabagi et al. 2005).
The catalysis, NAD interactions, evolution, and conformational changes of ADHs have been investigated (Eklund and Bränden 1979; Danielsson et al. 1994; Persson et al. 1994), using three-dimensional structures from the horse liver and focusing on the analysis of the differences among the enzymes of distinct species. Other studies have considered plant Adh evolution (Chang and Meyerowitz 1986; Gaut and Clegg 1993; Perry and Furnier 1996; Morton et al. 1996; Miyashita et al. 1998; Charlesworth et al. 1998; Gaut et al. 1999; Koch et al. 2000; Lin et al. 2001). Plant Adh transcription has been demonstrated to increase by environmental stresses such as low oxygen levels, dehydration, low
59 activation of the fermentation pathway compensates the decrease of the Tricarboxylic Acid Cycle function and of oxidative phosphorilation, regenerating NAD+ and producing energy. Phylogenetic studies indicated two or three isozymes, sometimes more than three, in all flowering dicot and monocot plant species, except in Arabidopsis, where a single Adh locus is found. Differences in Adh1 alleles specific activity were detected in maize, while different patterns of tissue-specific expression were observed in the Adh1 and Adh2 loci (Gaut and Clegg 1993; Gaut et al. 1999). However, no consideration was given to the relationship between structure and evolution, since there was no three-dimensional model of the plant alcohol dehydrogenases available. Gaut et al. (1999) assumed that the horse and plant ADH structures were similar and mapped some amino acid replacements of plant onto the horse secondary structure. Actually, there is a powerful method to model protein three-dimensional (3D) structures, which makes easier to locate the amino acid residues important to the functional diversification of enzymes and predict substrate preferences. This method (comparative protein structure modeling) estimates the 3D structure of a given protein sequence based on its alignment to one or more templates (Martí-Renom et al. 2000).
Experimental studies have shown the ADH involvement in additional metabolic pathways in plants, indicating putative distinct enzymatic functions during tobacco’s pollen tube growth (Bucher et al. 1995) and seed storage (Zhang et al. 1994, 1995a, 1995b, 1997), in potato’s pollinated pistils (van Eldik et al. 1997) and in Petunia’s seed detoxification (Garabagi et al. 2005).
60 We recently proposed the first plant ADH three-dimensional model using Arabis
blepharophylla data (Thompson et al. 2007), obtaining evidence for variation in the
subunit-subunit interacting segment, active site and the loop around the second zinc atom. The present work provides 16 other 3D structures, which are considered together with the first described especially in relation to their electrostatic and pI properties. The amino residues theoretically important to the functional divergence among the Poaceae, Brassicaceae, Fabaceae, and Pinaceae modeled ADHs were indicated, as well as those between ADH subtypes, and their position in the 3D structure evaluated to contribute to the elucidation of their functional divergence and molecular evolution.
61 Source of the data and sequence alignment
A total of 16 alcohol dehydrogenase sequences were retrieved from the National Center of Biotechnology Information (NCBI) and added to a previous one reported by Thompson et al. (2007). They are listed in Table 1. As indicated there, the 12 species from which they were isolated could be classified in four botanical families. Representatives from ADH1, ADH2 and ADH3 proteins were considered. The ClustalW program (Jeanmougin et al. 1998) was used to perform the alignments, which were inspected and manually changed when necessary using GeneDoc 2.6 (Multiple Sequence Alignment Editor & Shading Utility) (Nicholas and Nicholas 1997).
Modeling
Three-dimensional structures for the 17 ADH enzymes were built using the Equus
caballus liver form (PDB code 1N8K) as a template, obtained through Blastp (Altschul et
al. 1990, 1997). Its structure has been solved to a 1.13 Å resolution (Rubach and Plapp 2003). The ClustalW program (Thompson et al. 1994) was employed to perform the amino acid sequence alignments, using the BLOSUM62 matrix (Henikoff and Henikoff 1992) for scoring. The penalties for gap opening and gap extension were 10.0 and 0.2, respectively. The GeneDoc 2.6 program (Nicholas and Nicholas, 1997) was used to plot the percent identity of the sequences and manually adjust the alignment. The plot is created by sorting the data to be plotted into ascending order. For each data point the fraction of data points which have the same or a smaller value is computed. The data is then compressed to
62 eliminate multiple points with the same value. The highest value is retained during the compression.
The protein models were obtained through MODELLER 8v2 (Sali and Blundell 1993), which implements an approach to comparative protein structure modeling by satisfaction of spatial restraints. The best model was selected using PROCHECK (Laskowski et al. 1993) and VERIFY-3D (Lüthy et al. 1992). The PROCHECK program calculates the stereochemical parameters of the main and side-chains, the residues in the most favored regions, bond lengths and the angle’s standard deviation. VERIFY-3D evaluates the compatibility of a 3D model with the amino acid sequence considered using a 3D profile. Each residue position in the 3D model is characterized by its location and environment (alpha, beta, loop, polar, nonpolar, etc), and it is represented by 20 numbers in the profile. These numbers are called 3D_1D scores. The residue environments are defined by three parameters: the residue area that is buried, the fraction of side-chain area that is covered by polar atoms (O and N), and the local secondary structure. If the model is correct, the sum of the 3D profile scores is high, preferentially above zero. The protein signatures were obtained using a database of protein domains known as PROSITE (Gattiker et al. 2002). The protein volumes and surface areas were calculated according to the Richards' Rolling Probe Method (Richards 1974, 1977), using the 3V program (Voss Volume Voxelator) (Voss 2006), with a 1.5 Å probe radius and a high grid resolution (0.5 Å). The theoretical isoelectric point and molecular weight were obtained using the ExPASy Tools available at http://ca.expasy.org/tools/pi_tool.html (Gasteiger et al. 2005). Koch et al. obtained a value (5.81) not significantly different from our results (5.65) for the
63 mean square deviation (RMSD) between the template and the model and also to compute the electrostatic potential using the Coulomb method, as well as to draw all the figures and to generate the molecular surface. The nicotinamide-adenine-dinucleotide (acidic form) and two zinc atoms present in the PDB 1N8K code were located in the modeled three- dimensional structures using its fitting tool. The theoretical models are available upon request.
Functional divergence analysis
The amino acid residues responsible for the functional divergence of the plant ADHs were predicted based on site-specific profiles in combination with suitable cut-off values derived from the posterior probability of each comparison, using Gu’s (2001) methodology, as in our previous analysis (Thompson et al. 2007). It is known that functional changes are highly correlated to variations in the evolutionary rates occurring during a certain period of time. Therefore, the identification of the residues submitted to this process in our material were evaluated by finding sites with very different patterns (e.g., very few changes in one cluster but many in the others).
The site-specific profile to identify responsible amino acid sites uses a Qk to be the posterior probability that site k is in state S1 (0 ≤ Qk ≤ 1). A large Qk indicates a high possibility that the functional constraint (or the evolutionary rate) of a site is different between two clusters. We used three cut-off values, equal to or above respectively 0.80, 0.85, and 0.90.
64
Results
Sequence alignment and modeling
The results obtained with the multiple alignments are presented in Fig. 1, and they show high similarity among the sequences. The degree of identity between the sequence of the selected template and the models was around 48%. In general, the number of gaps in the template’s primary sequence is very low (Fig. 1), so it does not significantly affect the comparative molecular modeling. The inserted region of the alignment (Fig. 1 – positions 75 up to 83), which do not have an equivalent segment in the template, was modeled in the context of the whole molecule, using its primary sequence alone. The percent identity of the sequences is presented in Fig. 2. All target proteins have the signature of the zinc alcohol dehydrogenase family, which has a consensual pattern corresponding to G-H-E- X(2)-G-X(5)-[GA]-X(2)-[IVSAC]. Ten models were initially created, and they were considered using PROCHECK and VERIFY-3D, as well as the root mean square deviations (RMSD).
The stereochemical parameters used to verify the quality of the models are listed in Tables 1S-3S (Electronic Supplementary Information). In relation to the main-chain, of the six parameters considered, only the average of bad contacts per 100 residues show some differences among the families considered, the number in the Brassicaceae being higher (8.17) than in the Poaceae, Fabaceae, or Pinaceae (7.57, 6.00, and 7.00), respectively. The figures for the side-chains show better fit (smaller numbers) in relation to Chi1 gauche (-), Chi1-trans, and Chi-pooled SD for the Brassicaceae, the Poaceae model also comparing favorably in relation to the others for Chi-1 gauche (+), Chi1 pooled SD and Chi-2 trans. The mean of percentage of amino acid residues in most favored regions according to the
65 not significant since all results above 90% are considered of good quality. No model value was lower than 90.8%, confirming the excellent quality of the initial models.
Taken together, these data suggest that the models were stereochemically valid. It is important to observe that the G factor measures how “normal” is a given stereochemical property, considering the torsion angles and the bond lengths in the main chain. Therefore, when applied to a specific residue, a low G factor indicates that the property corresponds to a low probability of conformation. A G factor value smaller than -1.0 could indicate geometry problems. In this work, all G factor results are near -0.1 (Tables 1S-3S, Supplementary Information). Observing the VERIFY_3D (Figures 1S-3S, Supplementary Information) results, we can see that the sum of 3D profile scores is high in all cases. The region near the 301 amino acid residue, however, shows the smaller 3D_1D average scores for all botanical families, which means that this is most likely the area with the higher number of structural problems. In a general way, the graphics show a similar pattern.
Number of residues, molecular weight, surface area, and volume
Information concerning these four variables is shown in columns 3-6 of Table 2.
Brassica oleraceae (1BRAOLE) has a reduced number of amino acids (350), conditioning
also lower values for the molecular weight and volume. The opposite occurs in 1ZEAMAY which presents the highest number of residues (388). No clear differences in relation to these variables were observed in the ADHs of different botanical families.
66 Electrostatic and pI differences
The molecular surface of this protein is electrostatically polarized (Figs. 3-5). The Brassicaceae have the most acid ADH proteins when compared to the other families (Table 2), with Brassica oleraceae (1BRAOLE) having the most negative pI (5.47) value, followed by Arabis blepharophylla (2ARABLE; 5.65), Arabis griffthiana (1ARAGRI; 5.69), and Arabis parishii (1ARAPAR; 5.88). The regions of the active site, the second zinc atom, and of the subunit-subunit interacting segment (middle portion, upper and lower right region of the figures, respectively) show the greatest differences (Fig. 3). These proteins have a pI value significantly different from those of the Leavenworthia proteins (2LEAST and 3LEAST), which show pI values equal to 6.37 and 6.40, respectively.
Considering now the Poaceae group (Table 2 and Fig. 4), it is seen that the ADH1 forms 1HORVUL and 1ORYSAT (pI 6.28 and 6.20; nos. 1 and 4 in the Figures) are more basic than the ADH2 forms of the same species (respectively 5.52 and 6.04; nos. 2 and 5 in the Figure), the same occurring in Zea mays (6.43 and 5.72, nos. 6 and 7 in the Figure). The most significant differences in electrostatic potential is in the region near the second zinc atom and in the subunit-subunit interacting segment (upper and lower right, Fig. 4), a smaller contrast being observed in the active site region.
The pI values for the Fabaceae are not much different (Table 2). However, there is a different concentration of negative charges between the models, the subunit-subunit segment of Lotus corniculatus (Fig. 5.1) showing a clear difference from the other two (Figs. 5.2 and 5.3). The Pinus banksiana model have a pI value of 5.91 (Table 2), and the protein has the negative charges concentrated in the active site region (Fig. 5.4).
67 Sites showing Qk values above 0.8 and therefore suggestive of being associated with functional divergences are listed in Table 3 for the comparisons involving different botanical families (60 sequences considered); while in Table 4 the comparisons are between the ADH1 and ADH2 forms. Data related to ADH3 could not be used for this analysis because the number of sequences available was less than those needed for statistical comparisons (Gu and Vander Velden 2002).
Concentrating our attention to the comparisons which yielded Qk ≥ 0.9 only, we see that in the Brassicaceae vs. Fabaceae contrast, three residues which occur in loops (133, 303, 310) (Table 4S) show different amino acid conservation, the same being true for two others (315, 337) that are located in helices (Table 4S). Positions 133 (Gly), 303 (Ser), and 337 (Gly) are conserved within Brassicaceae, but highly divergent in Fabaceae (133: Asn, Gly, Ser; 303: Asn, Ser, Lys; 337: Asn, Leu, Ser, Gly) (Table 3). The three Fabaceae modeled show variability in position 303 only (Table 4S). In the Poaceae vs. Fabaceae comparative analysis, two amino acid residues of loop regions (118, 133) and one in the helix secondary structure (236) (Table 4S) should be considered; while in the Fabaceae vs. Pinaceae comparison the amino acids to be distinguished are 131 and 133 (loop) and 337 (helix) (Tables 3 and 4S). The Poaceae ADHs present conservation in residues 118 (Asp) and 133 (Gly), and the six Fabaceae in residue 236 (Phe) (Table 3). Considering the three Fabaceae ADH modeled, position 118 is variable in the Poaceae vs. Fabaceae, and position 131 in the Fabaceae vs. Pinaceae comparisons (Table 4S).
As presented in Tables 4 and 5S, both in the functional divergence analysis and in the models, amino acids that show different rates of change between Poaceae’s ADH1 and
68 ADH2 are nos. 234 and 263 (in helices) and 329 (loop), ADH2 being conserved for all of them. In the Poaceae vs. Fabaceae comparison the Poaceae ADH1s exhibit differences in residues 263, located in helix and 329, in a loop.
A ribbon representation of one model of each botanical family showing sites identified as functional divergent (Qk ≥ 0.85) is presented in Figs. 6 and 7. The subunit- subunit interaction segment seems to be the region with the highest number of functionally important residues in Brassicaceae and Fabaceae (Figs. 6.1 and 7.1, respectively). In Fabaceae the amino acids forming helices and loops around the second zinc atom region are variable (Fig. 7.1). Amino acid changes near the same region distinguish the Poaceae ADH forms, as well as substitutions in the dimer interaction zone (Fig. 6.2). The same regions are fundamental for the diversification of Pinaceae ADHs (Fig 7.2). There are also some differences among all ADHs near the coenzyme region (in green).
Discussion
Alcohol dehydrogenase is an essential enzyme in the anaerobic metabolism, and it has been widely used as a molecular marker in plants due to its convenient size (2-3kb in length with a ~1000 nucleotide coding sequence, 10 exons, 9 introns) and low copy number. The enzyme is important primarily for responses to hypoxic conditions, when its expression is highly induced. Moreover, it has an important role in fruit ripening, seedling and pollen development (Small and Wendel 2000). Despite the large number of phylogenetic investigations performed, no extensive work correlating its sequence and structure in plants exists.
69 different specific activity, and distinct pattern of organ-specific gene expression (Schwartz and Laughner 1969; Freeling and Bennet 1985). An exchange of Tyr for Asp at residue 52, located in a helix structure in the Adh1-C allele, alters enzymatic properties by reducing the specific activity. Additionally, amino acid replacements changing the secondary structure were also reported (Gaut and Clegg 1993).
In humans, ADH is a cytosolic enzyme able to metabolize ethanol and a wide variety of substrates, including aliphatic alcohols, hydroxysteroids and lipid peroxidation products. Its catalytic properties are variable. The Adh2 gene may be present as Adh2*1,
Adh2*2, and Adh2*3 encoding for β1, β2, and β3 subunits, respectively, which differ by a
single nucleotide change. The enzyme containing the β1 subunit has high affinity and low capacity for ethanol, whereas the β2 and β3 forms show lower affinity and higher capacity. Additionally, the human tissues show measurable different Adh gene expressions (Gemma et al. 2006).
The proteins modeled in this work are composed by two domains and have a similar fold. The nucleotide binding domain is formed by a structural motif known as Rossmann fold (Lesk, 1995), consisting of parallel beta strands linked by alpha helices (Figs. 6 and 7, region of nicotinamide binding at lower right). The catalytic region containing residues involved in substrate binding has a zinc atom located deeply in the cleft formed between the two domains. There are divergent amino acid residues localized in three important regions (the loop around the zinc atom, an important cofactor for the enzyme’s function; the subunit-subunit interacting segment, responsible for the dimer formation; and the active site) which are probably submitted to functional diversification.
70 Zinc seems to be important for the catalysis and geometry stabilization of the active site. These two processes could be achieved by moderating the electrostatic potential near the substrate or by zinc acting as ligand during the enzyme’s catalysis (Baker et al. 2008). Thus the residues indicated as functionally divergent near the zinc atom region possibly have an impact on ADH function. Some residues located near the zinc atom region, such as 109 and 112, which were not previously discussed since they have 0.80 ≤ Qk ≤ 0.85, may be also candidates for future investigations. The same can be said of 313 that is related to the subunit-subunit interaction, and residues nos. 49, 62 and 178, present near the active site. The first helix, located in residues 49 up to 55 using 1HORVUL as reference, can accommodate large movements associated with the loop near the active site (Baker et al. 2008); consequently, amino acid no. 64 (loop) has high probability to contribute to these movements.
Clearly the modeled proteins show electrostatic potential differences in the